DE102015201688A1 - METHOD AND SYSTEMS FOR ANALYSIS OF HEALTH DATA - Google Patents

Abstract

Embodiments for generating a model that can identify one or more clusters in a health record are disclosed. An input belonging to a range of numbers is obtained. Each number in the range of numbers is representative of a cluster number in the health record. For a cluster, one or more first parameters of a distribution associated with the cluster are estimated. Thereafter, a threshold is determined based on the one or more first parameters. An inverse cumulative distribution of each of the one or more n-dimensional variables in the health record is determined. The one or more first parameters are updated to generate one or more second parameters based on the estimated inverse cumulative distribution. A model is generated for each number in the range of numbers based on one or more second parameters.

Description

BACKGROUND

The healthcare industry is an industry that handles the maintenance of various records from a birth certificate to a person's death certificate. These records may include, but are not limited to, medical diagnostic reports, health insurance reports, hospital records, etc. This record data can be used to generate a mathematical model that can identify / anticipate information, such as: A health condition of a patient and health insurance fraud, but not limited thereto. To generate the mathematical model, one or more patterns in the record data must be identified.

Data collection techniques may determine one or more patterns in the record data. Such patterns can be used to determine clusters in the record data. Clustering is the grouping of a set of records in the record data based on predefined features associated with the record set. Some of the commonly known clustering algorithms include k-means clusters, density-based clusters, centroid-based clusters, Gaussian merge models, and so on.

A Gaussian mixed model is a clustering technique which requires that the record data contain one or more components or clusters and that the data is normally distributed in each cluster (eg, Gaussian distribution). To train the Gaussian mixed model, an input associated with a cluster number present in the record data is received from a user. As described above, the data in each cluster is normally distributed. Parameters such as averaged or covariant distribution of each cluster can be estimated using the expectation maximization algorithm. In one embodiment, the expectation maximization algorithm includes determining a probability that a data point or record corresponds to a cluster. The probability is maximized and the parameters of the distribution leading to the maximized probability are selected. The selected parameters are used to generate the Gaussian mixed model.

Since it is assumed that the data in the clusters are normally distributed, the Gaussian mixed models can not be applied to scenarios where data is not normally distributed.

SUMMARY

In accordance with the embodiments presented herein, a method for generating a model that can identify one or more clusters in a health record is provided. The method includes receiving a numerical range input by one or more processors. Each number in the range of numbers is representative of a cluster number in the health record. For a cluster in the cluster count, one or more first parameters of a distribution associated with the cluster are estimated. An inverse cumulative distribution of each of one or more n-dimensional variables in the health record is determined based on a threshold and a cumulative distribution of each of the one or more n-dimensional variables. The one or more first parameters are updated to generate one or more second parameters based on the estimated inverse cumulative distribution, wherein the update is performed using an expectation maximization algorithm. Finally, the model is generated for each number in the range of numbers based on one or more second parameters associated with each cluster in the cluster number.

In accordance with the embodiment presented herein, a system is provided for generating a model that can identify one or more clusters in a health record. The system, which includes one or more processors, is configured to receive input associated with a range of numbers. Each number in the range of numbers is representative of a cluster number in the health record. Furthermore, one or more processors are configured to estimate one or more first parameters of a distribution associated with a cluster of a cluster number. The one or more processors are further configured to determine a threshold based on the one or more first parameters. An inverse cumulative distribution of each of the one or more n-dimensional variables in the health record is estimated based on the threshold and cumulative distribution of each of the one or more n-dimensional variables by one or more processors. The one or more processors are for updating the one or more first parameters configured to generate one or more second parameters based on the estimated inverse cumulative distribution, wherein the update is performed using an expectation maximization algorithm. The one or more processors are configured to generate the model for each number in the range of numbers based on one or more second parameters associated with each cluster in the cluster number.

In accordance with the embodiment illustrated herein, a computer program product is provided for use with a computing device. The computer program product includes a non-transitory computer-readable medium. The non-transitory computer-readable medium stores computer program code for generating a model that can identify one or more clusters in a health record. The computer program code may be executed by one or more processors in the computing device to receive an input associated with a range of numbers. Each number in the range of numbers is representative of a cluster number in the health record. For a cluster in the cluster count, the computer program code may be executed to estimate one or more first parameters of a distribution associated with the cluster. Thereafter, a threshold is determined based on the one or more first parameters. An inverse cumulative distribution of each of one or more n-dimensional variables in the health record is determined based on a threshold and a cumulative distribution of each of the one or more n-dimensional variables. The one or more first parameters are updated to generate one or more second parameters based on the estimated inverse cumulative distribution, wherein the update is performed using an expectation maximization algorithm. Finally, the model is generated for each number in the range of numbers based on one or more second parameters associated with each cluster in the cluster number.

In accordance with the embodiments presented herein, a method is provided for stratifying one or more patients into one or more categories based on medical record data associated with each of the one or more patients. The medical record data includes a measurement of one or more physiological markers of each of the one or more patients. The method includes receiving a numerical range input by one or more processors. Each number corresponds to a category number in the medical record data. Each category corresponds to a medical condition associated with each of the one or more patients. For a category in the category number, one or more first parameters of a distribution associated with the category are estimated. An inverse cumulative distribution of the one or more physiological markers is determined based on a threshold and a cumulative distribution of each of the one or more physiological markers. The one or more first parameters are updated to generate one or more second parameters based on the estimated inverse cumulative distribution. The update is performed using an expectation maximization algorithm. For each number in the range of numbers based on one or more second parameters associated with each category of the category number, a model is generated. A best model is selected from the model generated for each number in the range of numbers using the Bayes information criterion. The best model is deterministic of the number of categories in the medical record data. The best model stratifies each of the one or more patients listed in the medical record data into one or more categories.

BRIEF DESCRIPTION OF THE DRAWINGS

The attached drawings illustrate various embodiments of systems, methods, and other aspects of the disclosure. Any one of ordinary skill in the art will appreciate that the illustrated element boundaries (eg, fields, arrays, or other shapes) in the figures represent an example of the boundaries. In some examples, one element may be configured as multiple elements or multiple elements may be configured as one element. In some examples, an element represented as an internal component of one element may be implemented as an external component in another and vice versa. Furthermore, the elements are not necessarily to scale.

Various embodiments will be described below with reference to the accompanying drawings, which are provided for illustration and not limitation of scope in any way, wherein like reference numerals represent like elements. Show it:

1 a flowchart illustrating a method for generating a model that can identify one or more clusters in a multivariate dataset;

2 a flowchart illustrating the generation of the model according to at least one embodiment;

3 a block diagram of a computing device that may generate the model according to at least one embodiment; and

4 5 is a flowchart illustrating a method of stratifying one or more patients based on medical record data associated with the one or more patients according to at least one embodiment.

DETAILED DESCRIPTION

The present disclosure may best be understood by reference to the detailed figures and descriptions presented herein. Various embodiments are discussed below with reference to the figures. One skilled in the art will appreciate, however, that the detailed descriptions presented herein are merely exemplary with reference to the figures, as the methods and systems may extend beyond the described embodiments. For example, the teachings and requirements of a particular application may provide several alternative and appropriate approaches for implementing the operation of any of the details described herein. Therefore, any approach may extend beyond the particular implementation choices of the following embodiments described and illustrated herein.

References to "one embodiment," "at least one embodiment," "an example," "for example," etc., indicate that the embodiment (s) or example (s) so described have a particular feature, structure , Characteristic, property, element, or constraint, but not every embodiment or example necessarily has that particular feature, structure, characteristic, property, element, or constraint. Furthermore, the repeated use of "in one embodiment" does not necessarily refer to one and the same embodiment.

Definitions: For purposes of this application, the following terms are intended to have the same meanings as listed below.

"Multivariate dataset" refers to a dataset that contains observations about a p-dimensional variable. For example, "n" executions of the p-dimensional variable may constitute a multivariate dataset. For example, medical record data may include a measurement of one or more physiological parameters of one or more patients. Such medical record data is an example of a multivariate record.

"Health record" refers to a multivariate dataset containing data obtained from the healthcare industry. In one embodiment, the health record may correspond to patient data, hospital data, health insurance data, diagnostic data, and so on. In a scenario in which the health data corresponds to the patient data, the one or more physiological parameters correspond to the p-dimensional variable, and the recording number in the health data corresponds to the explanations.

"Gaussian Mixed Model (GMM)" refers to a mathematical model that can identify one or more clusters in the multivariate dataset. In one embodiment, the data values in each of the one or more clusters are normally distributed (eg, Gaussian distribution).

"Gaussian Copula Mixed Model (GCMM)" refers to a mathematical model that can identify one or more clusters in the multivariate dataset, with the data values in each of the one or more clusters distributed according to a Gaussian copula distribution are.

"Cumulative distribution" refers to a distribution function that determines the probability that a real valued random variable X with a given probability distribution will be found at a value less than or equal to x.

"Inverse cumulative distribution" refers to an inverse function of the cumulative distribution of the random variable X.

Cluster Mixing refers to a probability that a data value in the multivariate record belongs to different clusters. For example, the multivariate dataset contains two clusters. A probability that a data value in the multivariate record belongs to the first cluster is 0.6. Therefore, the probability that the data value belongs to the second cluster is 0.4. In one embodiment, the sum of the data value probability in each of the one or more clusters in the data set is one.

"Deferred variable" refers to an intermediate variable that is not obtained from the multivariate dataset. In one embodiment, the latent variable is determined based on the one or more parameters.

"Probability" is to be understood comprehensively and contains every probability calculation; Probability approximation using any type of input data, regardless of precision or lack of precision; any number, whether calculated or predetermined, that simulates a probability; or any process step that has an impact on the use or finding of some data with a probability relationship.

As described, the Gaussian merge models are used to determine one or more clusters in a data set. To determine the clusters, the Gaussian mixed models assume that data points in a cluster are normally distributed. In one embodiment, in most of the applications, the data points may not normally be distributed. Therefore, the Gaussian blended models may not be able to accurately predict the clusters in the data set.

worin

y_i:

die inverse kumulative Verteilung der p-dimensionalen Zufallsvariabel x ist;

p:

die Dimensionenzahl der zufälligen Variabel ist;

π_g:

der Clustermischanteil g in Bezug auf andere Cluster in dem multivariaten Datensatz ist;

ψ_j(y_i,j):

die Randdichte von GMM entlang der Dimension j^th ist;

G:

die Clusterzahl in dem multivariaten Datensatz ist;

μ_g:

der Mittelwert der Gauß'schen Copula-Mischkomponente g ist;

Σ_g:

die Kovarianzmatrix der p-dimensionalen Variabel x (die eine Kovarianz zwischen dem einen oder den mehreren Clustern darstellt) ist; und

ϕ(y_i|μ_g, Σ_g):

eine multivariate Gauß'sche Verteilung der Datenwerte in einem Cluster g mit dem Mittelwert μ_g und der Varianz Σ_g ist.

In one embodiment, a Gaussian Copula Mixed Model (GCMM) is another mathematical model used to identify one or more clusters in a multivariate dataset. In one embodiment, the multivariate dataset may include data values of one or more p-dimensional variables. Each data value of each of the one or more p-dimensional variables may be part of a cluster in a multivariate data set. In one embodiment, the GCMM assumes that the data values in the cluster are derived from a Gaussian copula distribution. In one embodiment, copula corresponds to a multivariate probability distribution for which an edge probability of each variable is uniformly distributed. In one embodiment, the copula is used to describe the dependency between one or more p-dimensional variables in the data set. A typical Gaussian copula mixing model (GCMM) is represented by the following equation:

wherein

y _i :

is the inverse cumulative distribution of the p-dimensional random variable x;

p:

the dimension number of the random variable is;

π _g :

the cluster mix fraction g is relative to other clusters in the multivariate dataset;

ψ _j (y _{i, j):}

the edge density of GMM is along the dimension j ^th ;

G:

is the cluster number in the multivariate dataset;

μ _g :

the mean value of the Gaussian copula mixing component is g;

Σ _g:

the covariance matrix of the p-dimensional variable x (representing a covariance between the one or more clusters); and

φ (y _i | μ _g , Σ _g ):

is a multivariate Gaussian distribution of the data values in a cluster g with the mean μ _g and the variance Σ _g .

To determine the cluster number in the multivariate data set and to classify each data value of the one or more p-dimensional variables, a GCMM is generated. Generation of a GCMM in an embodiment of the disclosure has been described in connection with 1 described. 1.

1 is a flowchart 100 , which is a method of creating a model that can identify one or more clusters in a multivariate dataset. In one embodiment, the model is a Gaussian Copula Mixed Model (GCMM).

At step 102 an input is received from a user. In one embodiment, the input corresponds to a range of numbers. In one embodiment, the number range corresponds to a GCM model number to be generated. In addition, in one embodiment, each number in the range of numbers corresponds to a probable cluster number that may be present in the multivariate dataset. For example, if the user inputs the range as 1 to 3, three GCM models are generated for each number in the range (ie, 1, 2, and 3). Furthermore, each number (ie, 1, 2, and 3) represents the cluster number in the multivariate dataset. For example, for the number 3 in the range of numbers, the multivariate dataset may contain three clusters. In one embodiment, the GCM models generated for a particular number in the range of numbers may identify that particular cluster number in the multivariate dataset. For example, the GCM model generated for the number 3 in the range of numbers may identify three clusters in the multivariate dataset.

In addition, the multivariate record is obtained from the user. The multivariate dataset contains data values belonging to a p-dimensional variable in the multivariate dataset. In the following, the term data value is used synonymously with execution. For purposes of the present description, there are n executions of the p-dimensional variable in the multivariate dataset.

π_g:

die Mischanteile des einen oder der mehreren Cluster ist;

Σ_g:

die Kovarianz zwischen dem einen oder den mehreren Clustern ist;

G:

die Clusterzahl in dem multivariaten Datensatz ist;

y (0) / i,j:

die inverse kumulative Verteilung der p-dimensionalen Variabel entlang der Dimension j^th ist; und

κ⁽⁰⁾:

Max(μ_g,j), worin μ_g,j dem Mittelwert der Clusterverteilung g entlang der Dimension j^th entspricht.

At step 104 one or more parameters associated with a cluster of one or more clusters are estimated. Before determining the one or more parameters, a number is sequentially selected from the range of numbers. In one embodiment, the number of cluster numbers in the one or more clusters. For each cluster in the one or more clusters, the one or more parameters are determined. In one embodiment, the one or more parameters may include, but are not limited to, a merge portion of one or more clusters, an average of the distribution of the cluster (ie, Gaussian copula blend), a covariance between the one or more clusters. In one embodiment, the one or more parameters are determined randomly. In an alternative embodiment, the one or more parameters are determined using the k-means clustering algorithm. In one embodiment, the k-means clustering algorithm estimates the one or more parameters based on the following constraints: π _g > 0 (2) Σ G / g = 1π _g = 1 (3) Σ _{g is} positive and definite (4) δ _i = Min _{g, j} | y (0) / i, j - 2κ ⁽⁰⁾ ([[Σ (0) / g + I] ^-1 Σ (0) / gI) _j | (5) wherein

π _g :

the blend portions of the one or more clusters;

Σ _g:

the covariance between the one or more clusters is;

G:

is the cluster number in the multivariate dataset;

y (0) / i, j:

is the inverse cumulative distribution of the p-dimensional variable along the dimension j ^th ; and

κ ⁽⁰⁾ :

Max (μ _{g, j} ), where μ _{g, j} corresponds to the mean value of the cluster distribution g along the dimension j ^th .

One of ordinary skill in the art will understand that the scope of the disclosure is not limited to estimating the one or more parameters using the k-means clustering algorithm. In one embodiment, any other technique such as a decision tree and Gaussian mixed model may be used to estimate the one or more parameters.

worin

Γ:

der Schwellenwert ist;

S^(t) = Σ G / g=1z (t-1) / igΣ (t) / g (7) worin

z_ig:

einer latenten Variabel entspricht; und

m^(t):

die Summe aller Elemente von S^(t) ist.

At step 106 a threshold is determined based on the one or more parameters. In one embodiment, the following equation is used to determine the threshold:

wherein

Γ:

the threshold is;

S ^(t) = ΣG / g = 1z (t-1) / igΣ (t) / g (7) wherein

_Zig :

corresponds to a latent variable; and

m ^(t) :

is the sum of all elements of S ^(t) .

In one embodiment, the latent variable corresponds to an intermediate variable that is not obtained from the multivariate dataset. In one embodiment, the latent variable is determined based on the one or more parameters. The determination of the latent variable will be described later in an embodiment of the disclosure.

y_ij = Max(y_ij, Γ) (9) worin

y_ij:

die inverse kumulative Verteilung der p-dimensionalen Variabel entlang der Dimension j^th ist; und

σ (t) / g,jj:

j^th das diagonale Element der Kovarianzmatrix des g-ten Clusters ist.

At step 108 are an inverse cumulative distribution of the p-dimensional variable based on the threshold (which is used in step 106 was determined) and the cumulative distribution of the p-dimensional variable. In one embodiment, the following equations are used to determine the inverse cumulative distribution:

y _ij = Max (y _ij , Γ) (9) wherein

y _ij :

is the inverse cumulative distribution of the p-dimensional variable along the dimension j ^th ; and

σ (t) / g, jj:

j ^{th is} the diagonal element of the covariance matrix of the gth cluster.

In one embodiment, threshold Γ is a low limit on the inverse cumulative distribution of the p-dimensional variable. If z. For example, if the particular value of the inverse cumulative distribution y _{ij is} smaller than the threshold Γ, the threshold Γ is selected as the value of the inverse cumulative distribution y _ij .

One of ordinary skill in the art will understand that initially, when the one or more parameters are estimated using the k-means algorithm, the inverse cumulative distribution is determined based on the initial one or more parameters. In addition, based on the initial estimate of the inverse cumulative distribution, an initial probability is determined. In one embodiment, the initial probability of a probability corresponds to the initial one or more parameters being deterministic for the GCM model. In one embodiment, the initial probability is determined using the following equation:

At step 110 the latent variable is calculated based on the one or more parameters and the inverse cumulative distribution of the p-dimensional variable (which in step 108 determined). In one embodiment, the latent variable is determined using the following equation:

At step 112 the one or more parameters are updated based on the determined latent variable. In one embodiment, the one or more parameters are determined using the following equations:

At step 114 An updated probability is determined based on the updated one or more parameters. In one embodiment, the updated probability is determined using the following equation:

L^(t+1):

die aktualisierte Wahrscheinlichkeit ist, die durch Verwenden des einen oder der mehreren aktualisierten Parameter bestimmt wurde;

L^(t):

die Wahrscheinlichkeit ist, die in der vorherigen Iteration bestimmt wurde; und

∊:

der vordefinierte Schwellenwert ist.

At step 116 a check is made to determine if a difference between the updated probability and the previous probability is less than a predetermined threshold. In one embodiment, the previous probability corresponds to a probability determined in the previous iteration. For example, during the first iteration of the method, the probability for the first iteration (t = 1) is compared to the initial probability determined using Equation 10. Similarly, in each iteration, the probability is compared by using the updated one or more parameters for that iteration with the probability determined in the previous iteration. In one embodiment, the following equation is used to perform the test: L ^{(t + 1)} - L ^(t) <ε (16) wherein

L ^{(t + 1)} :

is the updated probability determined by using the one or more updated parameters;

L ^(t) :

the probability that was determined in the previous iteration; and

ε:

is the predefined threshold.

If at step 116 it is determined that the difference is greater than the predetermined threshold, the steps 106 to 116 repeated. If, however, at step 116 it is determined that the difference is less than the predetermined threshold, the updated one or more parameters are considered to be the model parameters.

u_ip:

die kumulative Verteilung der p-dimensionalen Variabel ist;

C:

die Kopulafunktion (von Gleichung 1 repräsentiert) der p-dimensionalen Variabel ist;

f_J(x_ij):

die gemeinsame Verteilung der p-dimensionalen Variabel ist;

υ:

der Vektor des einen oder der mehreren Parameter ist.

At step 118 a model is generated based on the updated one or more parameters. In one embodiment, the following equation represents the model: GCM - Model = Π n / i = 1Σ G / g = 1π _g Π n / i = 1 [C ((u _i1 , ..., u _ip ) | υ) Π p / j = 1 f _J (x _ij ) ] (17) wherein

u _ip :

is the cumulative distribution of the p-dimensional variable;

C:

the copula function (represented by Equation 1) is the p-dimensional variable;

f _J (x _ij ):

the common distribution of the p-dimensional variable is;

υ:

is the vector of the one or more parameters.

In one embodiment, the steps become 104 to 118 for each number in the number range repeated to generate the model for each number in the number range. Therefore, the number of models generated corresponds to the number range.

v ^:

der eine oder die mehreren aktualisierten Parameter ist, die zum Erzeugen des Modells in Schritt 118 verwendet werden;

L:

die Wahrscheinlichkeit ist (mithilfe der Gleichung 15), die für den einen oder die mehreren aktualisierten Parameter, die zum Erzeugen des Modells in Schritt 118 verwendet werden, geschätzt wird;

ρ:

die Zahl der freien Parameter ist; und

n:

die Zahl der Datenwerte oder Ausführungen ist.

At step 120 a best model is selected from the model generated for each number in the range of numbers. In one embodiment, the best model is selected using the Bayes Information Criterion (BIC). To determine the best model, a score is determined for each model generated for the numbers in the range of numbers. In one embodiment, the following equation is used to determine the score: BIC - score = 2logL (v ^ | (u _i1 , ..., u _ip )) - ρlogn (18) wherein

v ^:

is the one or more updated parameters that are used to generate the model in step 118 be used;

L:

the probability is (using Equation 15) that the one or more updated parameters used to generate the model in Step 118 be used is estimated;

ρ:

the number of free parameters is; and

n:

the number of data values or executions is.

In one embodiment, the free parameters correspond to parameters that are not dependent on the one or more parameters or the multivariate data set. The free parameters are determined independently. In one embodiment, the number of free parameters for p-dimensional data and G-clusters is determined using the following equation: ρ = (G-1) + Gp + Gp (p + 1) / 2 (19)

In one embodiment, the model with the best BIC score is selected as the best model. Further, in one embodiment, the number (from the range of numbers) for which the best model is generated corresponds to the cluster number present in the multivariate dataset. If the number range z. 1 to 3, three models are generated, one for each number, i. H. 1, 2, and 3. Also, if the model generated for the number 2 has the maximum BIC score, the second model corresponding to the number 2 is selected. In addition, in this case, the cluster number present in the multivariate dataset is two.

One of ordinary skill in the art will understand that in step 120 certain cluster number is an estimate of the cluster number present in the multivariate dataset. In one embodiment, the multivariate dataset may include more than the estimated number of clusters.

In one embodiment, the models generated for each number in the number range are mixed models. In one embodiment, the hybrid model corresponds to a probabilistic model that can identify one or more clusters in the multivariate dataset. After selecting the best model, the best model is used to categorize each data point (execution of the p-dimensional variable) in the multivariate record in the one or more clusters.

In one embodiment, this corresponds to the flowchart 100 described method an expectation maximization algorithm (EM). Each iteration of the EM algorithm alternates between performing a set of expectation steps (E) that produce a dependency on the expectation of the estimated log likelihood, using the current estimate for the parameters (determination of the latent variables) and a set of maximization steps (M), which calculate the parameters that maximize the expected log likelihood of the E-steps. In one embodiment, the steps correspond 106 . 108 and 110 the E-steps of the EM algorithm during the steps 112 . 114 and 116 correspond to the M-steps of the EM algorithm.

2 is a flowchart 200 depicting the generation of the model according to at least one embodiment. The flowchart 200 was related to 1 described. 1.

The multivariate dataset (by 202 shown) is obtained from the user. In addition, the number range (by 204 shown) received from the user. For example, the number range contains (1 (by 204a shown), 2 (by 204b shown) and 3 (by 204c shown)). As described above, each number corresponds to a possible cluster number in the multivariate dataset 202 is included. For example, for the number 1 (by 204a provided) provided that the multivariate dataset 202 only one cluster (ie cluster-1 (by 206 shown)) contains. In the same way, for the number 2 (by 204a provided) provided that the multivariate dataset 202 two clusters (ie cluster-1 (by 206 shown) and cluster 2 (by 208 shown). Furthermore, for the number 3 (by 204c shown) in the number range (by 204 provided) provided that the multivariate dataset 202 a third cluster (cluster 3 (by 210 shown)) in addition to the clusters 206 and 208 contains. For each number in the range of numbers, the EM algorithm is executed. In one embodiment, the EM algorithm estimates the one or more parameters of a blending model that can bundle the data points in the one or more clusters, where the cluster number is determined based on the number in the range of numbers. For example, the EM algorithm generated for cluster 1 (by 206 shown), the mixed model-1 212 containing the data values in the multivariate record 202 in cluster 1 (by 206 shown) can bundle. Similarly, the mixed model-2 (by 214 represented) for the number 2 (by 204b shown). The mixed model 2 (by 214 shown), the data values in the two clusters (ie cluster-1 (by 206 shown) and cluster 2 (by 208 shown)).

After generating the blended models for each number in the number range, a BIC score is calculated using Equation 18 (by 218 shown) for each mixed model. If the mixed model 2 (by 214 shown), for example, has the maximum BIC score, the mixed model-2 (by 214 shown) selected. Furthermore, and because the mixed model 2 (by 214 represented) for the number 2 (by 204b shown) in the number range (by 204 shown), the number of probable clusters in the multivariate dataset is 202 two. After selecting the mixed model-2 (by 214 shown) is the mixed model-2 (by 214 shown) for bundling (by 220 shown) of the multivariate dataset 202 used.

3 is a block diagram of a computing device 300 that can produce the model according to at least one embodiment. The computer device 300 has a processor 302 , a transceiver 304 and a memory 306 on. The processor 302 is with the transceiver 304 and the memory 306 coupled.

The processor 302 has appropriate logic, circuitry and interfaces and is capable of executing one or more instructions stored in the memory 306 for performing predetermined operations on the computing device 300 are stored, configured. The memory 306 may be configured to store the one or more instructions. The processor 302 can be implemented using one or more processor technologies known in the art. Examples of the processor 302 include, but are not limited to, an X86 processor, a RISC processor, an ASIC processor, a CISC processor, and any other processor.

The transmitter-receiver 304 sends and receives messages and data. Furthermore, the transceiver may receive the multivariate data set and the range of numbers from the user. Examples of transceivers 304 For example, but not limited to, an antenna, an Ethernet port, a USB port, or any other port that can be configured to receive and send data. The transmitter-receiver 304 sends and receives data and messages according to various communication protocols such as TCP / IP, UDP, and 2G, 3G, or 4G communication protocols.

The memory 306 stores a set of instructions and data. Some conventionally known memory implementations include, but are not limited to, RAM, read only memory (ROM), hard disk drive (HDD), and SD card. Furthermore, the memory contains 306 one or more instructions issued by the processor 302 to perform specific operations are executable. For a person skilled in the art it is obvious that one or more instructions are stored in memory 306 stored, the hardware of a computer device 300 allow to perform the predetermined operations. In one embodiment, the computing device is 300 to execute the flowchart 100 configured to generate the model that can identify the one or more clusters in the multivariate dataset.

In one embodiment, this can be done in the flowchart 100 described method for analyzing data from the healthcare industry. For example, patients can be stratified by disease detection patterns into disease risk profiles and treatment responses. The method may also be applied at different levels in the healthcare industry, such as: At individual patient level through analysis of the electronic health record (EMR) or at the hospital level (eg by identifying a group of patients at risk of being able to commit to health insurance fraud).

4 is a flowchart 400 10, which illustrates a method for stratifying one or more patients based on medical record data associated with the one or more patients, according to at least one embodiment.

At step 402 Medical records are obtained from the user. In one embodiment, the processor receives 302 the medical record data. In one embodiment, the medical record data includes, but is not limited to, measurements of various physiological markers of one or more patients, such as, but not limited to, age, blood pressure, serum cholesterol level, heart rate, and ST depression. In one embodiment, the physiological marker corresponds to the p-dimensional variable with age, blood pressure, serum cholesterol level, heart rate, and ST depression as different dimensions.

One of ordinary skill in the art will understand that the scope of the disclosure is not limited to the noted physiological markers. In one embodiment, various other physiological markers may be used.

At step 404 An input is received from the user who belongs to a range of numbers. In one embodiment, the processor receives 302 the input via the transmitter-receiver 304 , In one embodiment, the number range corresponds to a probable number of patients associated with the patient, which may be present in the medical record data. In one embodiment, the categories in the medical record data may correspond to a health status of one or more patients. For example, the one or more patients may be stratified into two categories, e.g. B. with heart disease and without heart disease.

At step 406 One or more parameters associated with a category of one or more categories are estimated. In one embodiment, the processor estimates 302 the one or more parameters similar to those in step 104 described way.

At step 408 an inverse cumulative distribution of the physiological markers is estimated. In one embodiment, the processor estimates 302 the inverse cumulative distribution. Before estimating the inverse cumulative distribution, the processor determines 302 the threshold, which is a lower bound on the inverse cumulative distribution of the physiological markers. In one embodiment, the threshold and inverse cumulative distribution may be as in the steps 106 respectively. 108 be determined described.

Based on the inverse cumulative distribution of the physiological parameters, an initial probability can be determined by using Equation 10.

At step 410 a latent variable is determined based on the inverse cumulative distribution of the physiological markers. In one embodiment, the processor determines 302 the latent variable. In one embodiment, the processor performs 302 the step 110 for determining the latent variable.

At step 112 the one or more parameters are updated based on the latent variable. In one embodiment, the processor is 302 configured to update the one or more parameters. At step 114 An updated probability is determined based on the updated one or more parameters. In one embodiment, the processor determines 302 the updated probability. At step 116 a check is made to determine if a difference between the updated probability and the previous probability is less than a predetermined threshold. If at step 116 determining that the difference is greater than the predetermined threshold 408 to 116 repeated. If, however, at step 116 it is determined that the difference is less than the predetermined threshold, the updated one or more parameters are considered to be the model parameters. At step 118 Furthermore, a model is generated based on the one or more updated parameters.

In one embodiment, the above steps are repeated for each number in the number range. In one embodiment, the number of generated models is equal to the total number present in the number range. Furthermore, at step 120 selected a best model from the models generated for the numbers in the number range. In one embodiment, the number from the number range for which the best model is selected represents the number of categories in the medical record data. If z. For example, if the best model for the number 2 is generated, the model may be the medical one Categorize recording data into two categories (eg patients with heart disease and patients without heart disease).

After generating the models and selecting the best model, the selected model is used to stratify the one or more patients into two categories, i. H. in patients with heart disease and patients without heart disease. In one embodiment, the best model categorizes the medical record data into the two categories based on the physiological parameters listed in the medical record data.

One of ordinary skill in the art will understand that in a scenario where new data is input by the user, the selected model may categorize the data into corresponding clusters based on the value of the p-dimensional variable. If z. For example, when physiological parameters of new patients are entered into the system, the model categorizes the new patients into one of two categories (eg, heart disease or no heart disease).

One of ordinary skill in the art will understand that the scope of the disclosure is not limited to stratifying the one or more patients and to the one or more categories. In one embodiment, similar medical data may be analyzed to make different conclusions. For example, health insurance data may be analyzed to determine health insurance amounts.

If the health data corresponds to health insurance data, the p-dimensional variable in the health insurance data may correspond to one or more insurance-related parameters such as the insured person's age, one or more physiological parameters of the insured person such as insurance premium paid by the insured person, amount of insurance and coverage. The one in the flowchart 100 and 400 The process described may be used to determine insurance amounts, recommended insurance contributions, and so on. In the same way, hospital data can be analyzed to help doctors make decisions and diagnoses.

The disclosed embodiments include numerous advantages. The estimation of the inverse cumulative distribution of the p-dimensional variable allows the use of the expectation maximization algorithm to generate the GCMM. Furthermore, the number of clusters present in the multivariate dataset is also estimated. In this way, the system becomes more dynamic and adaptable. Suppose the system receives an unknown multivariate dataset. The user may enter a range of numbers that he / she thinks should be the cluster number in the multivariate record. The system generates a model for each number and from the models thus generated, a best model is selected. The number from the range of numbers corresponding to the selected best model is representative of the cluster number in the multivariate dataset. This ability to estimate the cluster number makes the system adaptable. Furthermore, this adaptive system can be used to identify clusters in each multivariate dataset, such as health-related data.

The disclosed methods and systems as set forth in the foregoing description or any of the components may be implemented in the form of a computer system. Typical examples of a computer system include a general purpose computer, a programmed microprocessor, a microcontroller, a peripheral integrated circuit element, and other devices and arrangements of devices that may implement the steps that make up the method of the disclosure.

The computer system includes a computer, an input device, a display unit, and Internet. The computer further includes a microprocessor. The microprocessor is connected to the communication bus. The computer also has a memory. The memory may be Random Access Memory (RAM) or Read Only Memory (ROM). The computer system further includes a storage device that may be a hard disk drive or a removable drive such as a floppy disk drive, optical disk drive, and the like. The storage device may also be a means for loading computer programs or other instructions into a computer system. The computer system also has a communication unit. The communication unit allows the computer to connect to other databases and to the Internet via an input / output interface (I / O), allowing for transfer as well as receiving data from other sources. The communication unit may include a modem, an Ethernet card, or other similar device that allows the computer system to connect to databases and networks such as LAN, MAN, WAN, and the Internet. The computer system facilitates input from a user via the input devices that are accessible to the system via an I / O interface.

To process the input data, the computer system executes a set of instructions stored in one or more memory elements. The storage elements may also hold data or other information if desired. The storage element may be in the form of an information source or a physical storage element that is present in the processing engine.

The programmable or computer-readable instructions may include various instructions that instruct the processing engine to perform specific tasks, such as the steps that make up the method of the disclosure. The described systems and methods may also be implemented using only software programming or using only hardware or a varying combination of the two techniques. The disclosure depends on the programming language and the operating system used by the computers. The instructions of the disclosure may be written in all programming languages including, but not limited to, 'C', '++', 'Visual ++' and 'Visual Basic'. Furthermore, the software may be in the form of a collection of separate programs, a program module containing a larger program or portion of a program module, as discussed in the foregoing description. The software may also include modular programming in the form of object-oriented programming. The processing of input data by the processing engine may be in response to user commands, results of previous processing, or a request generated by another processing engine. The disclosure may also be implemented in various operating systems and platforms, including, but not limited to, 'Unix', 'DOS', 'Android', 'Symbian', and 'Linux'.

The programmable instructions can be stored and transferred to a computer-readable medium. The disclosure may also be embodied as a computer program product comprising a computer-readable medium, or any product that can implement the above methods and systems or the numerous different variations thereof.

Various embodiments of methods and systems for analyzing health data have been disclosed. It will be apparent, however, to one skilled in the art that changes may be made in addition to those described without departing from the inventive concepts herein. The embodiments are therefore not restrictive except in the spirit of the disclosure. Furthermore, when interpreting the disclosure, all terms are to be understood in the broadest possible sense in relation to the context. In particular, the terms "comprising" and "comprising" are to be understood as referring non-exhaustively to elements, components or steps that indicate that the related elements, components or steps are present or used or combined with other elements, components or steps which are not expressly described.

One of ordinary skill in the art will appreciate that the system, modules and sub-modules have been illustrated and explained to serve as examples and are not to be construed as limiting in any way. It will also be appreciated that the variations of the above-disclosed system elements or modules, as well as other features and functions or alternatives thereof, may be combined into many other different systems or applications.

One skilled in the art will appreciate that any of the foregoing steps and / or system modules may be replaced, rearranged, or removed, and that additional steps and / or system modules may be included, depending on the needs of a particular application. In addition, the systems of the foregoing embodiments may be implemented using a wide range of suitable methods and system modules and are not limited to any particular computer hardware, software, middleware, firmware, microcode, and the like.

The claims may include embodiments for hardware, software, or a combination thereof.

Claims (10)

A system for generating a model that can identify one or more clusters in a health record, the system comprising: one or more processors for: receiving, by one or more processors, an input associated with a range of numbers, each number in the range of numbers representing a cluster number in the health record; for a cluster in cluster number, for: estimating one or more first parameters of a distribution associated with the cluster; Determining a threshold based on the one or more first parameters; Estimating an inverse cumulative distribution of each of the one or more n-dimensional variables in the health record based on the threshold and a cumulative distribution of each of the one or more n-dimensional variables; Updating the one or more first parameters to generate one or more second parameters based on the estimated inverse cumulative distribution, wherein the update is performed using an expectation maximization algorithm; and for generating the model for each number in the number range based on one or more second parameters associated with each cluster in the cluster number. The system of claim 1, wherein the cumulative distribution of each of the one or more n-dimensional variables is determined from the health record. The system of claim 1, wherein the distribution associated with the cluster corresponds to a Gaussian copula distribution. The system of claim 1, wherein the expectation maximization algorithm further comprises determining a latent variable for the cluster based on the one or more first parameters and the inverse cumulative distribution of the one or more n-dimensional variables, and wherein the one or more first parameters updated based on at least the latent variables. The system of claim 1, wherein the expectation maximization algorithm further comprises determining a first probability that the one or more first parameters are deterministic for the model. The system of claim 5, wherein the expectation maximization algorithm further comprises determining a second probability that the one or more second parameters are deterministic for the model. The system of claim 6, wherein the one or more processors are further configured to compare the first probability and the second probability. The system of claim 7, wherein the model is generated using the one or more second parameters based on the comparison. The system of claim 7, wherein the threshold and the inverse cumulative distribution are updated using the one or more second parameters based on the comparison; and wherein the one or more second parameters are updated using the updated threshold and the updated inverse cumulative distribution based on the comparison, wherein the second probability is updated based on the updated one or more second parameters. The system of claim 1, wherein the one or more processors are further configured to select a best model from the model generated for each number in the range of numbers using the Bayes information criterion, the best model being deterministic of the category number in the health data is.

Images